Italy: Selected Issues
Author: International Monetary Fund
Publication Date:
16 Feb 2006
eISBN:
9781451819885
Language:
English
Keywords:
ISCR; CR; export; private sector; factors accounting; infrastructure investment; beginning-of-period product market regulation index; country dummy; Italy dummy; aggregate growth; rate of change; Exports; Tax incidence; Commodity markets; Labor taxes; Financial sector development; Global
Download PDF (1 MB)

This Selected Issues paper investigates the factors behind the deterioration in Italy’s international competitiveness. It concludes that the loss of competitiveness accumulated by Italian firms in recent years is mainly a consequence of weak long-term productivity performance. The paper explores the link between policies and growth. Specifically, it finds evidence that rigid product markets and a high tax burden on labor have been associated with slower growth in European regions. The paper also analyzes the role of fiscal policy and its implications for household consumption decisions.

Abstract

This Selected Issues paper investigates the factors behind the deterioration in Italy’s international competitiveness. It concludes that the loss of competitiveness accumulated by Italian firms in recent years is mainly a consequence of weak long-term productivity performance. The paper explores the link between policies and growth. Specifically, it finds evidence that rigid product markets and a high tax burden on labor have been associated with slower growth in European regions. The paper also analyzes the role of fiscal policy and its implications for household consumption decisions.

III. Regional Growth in the EU and Italy: Policies versus (Sectoral) Legacy1.

Core Questions, Issues, and Findings

  • What is the aim of the study? The chapter investigates factors behind recent growth performance in the EU and Italy using subnational (regional) data in cross-section and panel regressions for 1995–2004 (and various subperiods). Recent data advances permit a more consistent focus on the policy dimension, while recourse to subnational data augments testing power.

  • What are the main results? Within the EU, controlling for regional convergence, nationwide product market rigidities (as measured by the OECD) and the tax burden (notably on labor) were associated with negative growth in 1999–2002. In line with recent research, the short-term effect of labor market reforms on output was found to be negative in the EU. Within Italy, regional data show some effect of the “unfavorable” regional export specialization on growth in 1996–2004. While the short sample period means the results should be treated with caution, they are intuitively appealing, consistent with those of other studies, and relatively robust to different specifications.

  • What are the policy implications of the chapter’s findings? The results suggest scope for policies to overcome unfavorable initial conditions for growth, even in the medium-term horizon of 4–5 years. In particular, further product market liberalization and durable reductions in the tax burden may perceptibly help growth outcomes.

A. Introduction

1. The last few years have been marked by slow growth in the euro area. The consensus has been that this disappointing output performance reflected lingering structural weaknesses, possibly exacerbated by the “common” competitiveness problems brought on by the strong euro.2 The picture is, however, complicated by (i) substantial differences (and, at times, diverging experiences) across the European economies; (ii) uncertainty over the required policy response; and (iii) the role of exogenous factors, notably trade-related shocks.

2. Conditions in Italy have been particularly worrying, as poor growth has been accompanied by competitiveness problems. Unlike Germany, which has also suffered from slow growth, Italy’s measured productivity and competitiveness indicators have been deteriorating. The Italian growth/competitiveness nexus has been linked to broad-based factors (see Bank of Italy (2004–2005)), partly common to some other European countries (high tax burden, structural rigidities in labor and product markets, etc.) and partly Italy-specific (continued comparative wage and inflation differentials, rigid and slow bureaucratic and legal system, and relatively “outdated” export specialization). Still, despite numerous studies on various aspects of Italy’s performance, prioritization among the principal causes of weak growth - and the many proposed reforms - has remained an issue.3

3. This paper examines Europe’s and Italy’s recent growth experiences through regional data. While these data have already produced insights for both Europe and Italy, notably in terms of convergence, this study is different in some respects. First, the focus is on the medium term: 1995–2002 period for the Europe-wide dataset and 1995–2004 for Italy. While studies of growth determinants usually take a longer-term perspective, some important recent empirical work emphasizes medium-term growth experiences.4 The most recent period could be particularly revealing, as data variation reflects the key event of euro adoption, as well as structural and policy changes associated with integration and the widening of the EU. Second, the chapter employs a structured approach to disentangle national from some subnational factors, examining recent cross-country information on the quality of national policy environments, notably the OECD’s indices of market regulation. Third, with special reference to Italy-specific problems, regional data are used to gauge the role of competition from emerging markets compared to other measurable factors influencing concurrent output performance.

4. The remainder of the paper is structured in two principal parts, Europe-wide and Italy-specific. The first section examines the European NUTS-2 dataset, looking at pan-European regional convergence with and without country dummies, and then adding variables that summarize quality of national policy environments. The second section analyzes Italy’s regional performance in 1995–2004, and focuses on the interaction of common determinants of growth with factors related to external specialization. Some methodological caveats and data issues are elaborated in the appendix.

5. The main conclusions of the study are as follows:

  • Growth in European regions appeared to converge in 1995–2002, at a rate similar to that found in studies of other areas or periods. The sign and significance of the estimates generally holds even in shorter periods, and could potentially be used in filtering other determinants of growth.

  • Convergence disappears once country dummies are used as controls, which suggests that much of it has taken place among and not within countries - a result that could be reconciled with the view that the integration process has so far been largely a “top-down” phenomenon. However, country dummies are very coarse and difficult-to-interpret “controls.”

  • The OECD index of product market regulations (PMR) and the tax burden had a significantly negative association with growth in 1999–2002. At the same time, employment protection legislation (EPL) index had a positive (short-term) association with growth. Both results are broadly consistent with available evidence from other studies.

  • The country dummies in the augmented regressions have generally reflected differences in national growth rates. At the margin, the Italy country dummy influences cross-country results in a different way from the Germany country dummy, suggesting the relative importance of product market deregulation for Italy and of the tax burden on labor for Germany.

  • Italy-specific data suggest that, controlling for convergence and some other factors, the country’s unfavorable export structure explains only a small part of the growth variation across regions. Of course, this does not necessarily mean that unfavorable export specialization is unimportant for country-wide growth or export performance.

B. Europe-wide regional growth

Unconditional Convergence With and Without Country Dummies

6. Most studies of regional growth patterns have been based on the concept of convergence. The latter has been used as a metric for complementing and “filtering” other determinants of growth. Despite its wide use, the theoretical foundations of convergence are not universally agreed. A stylized one-sector neoclassical growth model with exogenous technological change predicts unconditional convergence (Barro and Sala-i-Martin (1995)), while weaker versions of the “convergence hypothesis” have stressed the crucial role of free trade and competition. However, there are also credible theoretical priors behind diverging dynamics, based, inter alia, on “increasing returns” (Krugman (1991)), or endogenous growth (Romer (1990)).

7. The empirical results of tests of the convergence hypothesis have - in Europe and elsewhere - reflected these theoretical ambiguities. Many studies point to long-term unconditional (absolute) convergence among and within industrialized countries. For Europe, the classic reference is Barro and Sala-i-Martin (1995), who, inter alia, found evidence of significant β- and σ- convergence in a group of European regions over 1950–1990, but their dataset excluded some European countries.5 Vamvakidis (2003) found steady convergence within 197 European regions through 2000. Other research has, however, detected unconditional divergence in a world-wide cross-country dataset, or even for certain periods within industrialized countries, and in Europe in particular. For example, Canova and Marcet (1995), and Quah (1997) argued that European regions are separating into four distinct clusters. More recently, Boldrin and Canova (2001) examined the period through 1996, and found neither convergence nor divergence with respect to per capita income in Europe, arguing, inter alia, that convergence weakened substantially after the late 1970s.6

Results

8. European regions exhibited significant unconditional beta convergence in 1995–2002, with standard estimated speed. Table 1 and Figure 1 present the main results in line with the basic convergence cross-section regression equation of Barro and Sala-i-Martin (1995):

(1/T)(lnyt−lny0)=α+βlny0+μ(1)

Where, for a given region, T is the number of years in the observed period (7 for the benchmark regression covering 1995–2002), and y is the level of PPP-adjusted per capita income at the beginning (y0) and at the end (yt) of the observed period.7 The two main samples are (i) 254 regions for EU-25 and (ii) 210 regions of EU-14 (EU-15 minus Luxembourg and two outlier capital cities (London and Brussels). Other country subsamples (Eurozone and others) were also tested. The convergence coefficient (β) is of the “expected” sign, and is equal to 1.3 percent for the EU-25 and 1.9 percent for EU-14, with the latter result close to the “standard” point estimate of 2 percent, reported for most cross- and within-country regressions by Barro and Sala-i-Martin (1995), including for their estimates in the European regions.8 The respective coefficient is confirmed in sign, but is much larger for panel regressions on annual data (especially the fixed-effects estimates), which is, again, consistent with the literature. In sum, these regressions suggest that, at least for the key period of 1995–2002, pan-European convergence did take place. Interestingly, however, fitting a non-linear trend (Figure 1) suggests that the convergence process tends to disappear among the richest regions.

Table 1.

Unconditional Convergence Regressions in the EU, 1996–2002. NUTS-2 Regions, Dependent Variable PPP Real per Capita Relative Changes

(log-form)

article image
Note: Regressions include a constant (not reported), t-statistics in parentheses.* denotes significance at percent and

at 1 percent; EU-14 excludes Luxemburg and twoinfluential capital cities (London and Brussels), EU-4 includes Italy, Spain, Germany and France.

Figure 1.
Figure 1.

EU-25, real PPP GDP per capita convergence, 1995–2002

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

9. Convergence also holds for most sub-samples over this timeframe, although, as expected, there is short-term variation due to cyclical factors. Still, even fairly short (3-year) subperiods yield quite strong and stable results on the direction and the strength of the pan-European convergence coefficients. As per Table 2, the convergence coefficient was again close to 2 percent for 3-year sub-periods for the EU-14, and for the first 3-year period (1996–98) for the EU-25. It declined to 1 percent for 1999–2002 for the EU-25, but remained highly statistically significant. The results for the individual years are much less stable (see Table 2), but most of the coefficients have the “expected” sign, including all those for the EU-25. Overall, this suggests that increasing the number of cross-section observations tended to smooth out short-term fluctuations around the medium-term trend.

Table 2.

Convergence Cross-section Regressions by Sub-period

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent

10. Including country dummies in the regional growth regressions provides little additional insight. As can be seen from Table 3, the convergence result disapears, as the β coefficient becomes statistically insignificant and even changes sign if all country-dummies are used as controls within the EU-25. This may suggest that much of the observed unconditional convergence has taken place between and not within countries.9 (At the same time, introducing country dummies is an imperfect way of disentangling within-country from between-country convergence). The signs of particular national dummies reported in Table 3 are in line with the basic cross-country relative growth rankings: Germany has the lowest growth, followed by Italy. These results are, however, of limited value as country dummies are hard-to-interpret “catch-all” variables.

Table 3.

Convergence Cross-section Regression with Country Dummies, 1996–2002 1/

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent

Includes country dummies for all 25 countries, but these are not reported for small countries.

11. Within-country results appear to confirm a predominantly cross-country nature of the recent convergence process. Table 4 shows that, of the large EU-5 countries, only Italy exhibited statistically significant β-convergence among its regions. Regions in Germany, France, and Spain also appeared to converge, but not significantly so. (Indeed, σ-convergence (not reported here) for the same period occured in Italy and France, but not in Spain or Germany). By contrast, the UK’s regions tended to diverge, although the extent and significance of the divergence would be smaller if London were removed from the sample.

Table 4.

OLS Convergence Cross-section Regressions within Large EU Countries, 1996–2002

article image
Note: t-statistics in parentheses

12. In any case, the unconditional convergence and/or country dummies can explain only a minor portion of regional and cross-country variation in growth rates. The R-squared in these simple regressions was generally in single digits, and improved only marginally with the addition of the country-dummies. To investigate the underlying causes of regional growth, a more complex, multifactor approach is needed.

Structural Regional Growth Determinants in the EU

13. A more structured way to gauge the determinants of growth in the EU requires a balancing act between key priors and available data. Existing cross-country literature on growth - which extends to regional issues - is fairly eclectic and considers a wealth of possible hypotheses, blending neoclassical growth determinants (based on capital and labor input accounting) with subsequent theoretical developments (i.e., human capital, innovation), institutions, and ad-hoc factors such as regional and country dummies (Barro and Sala-i-Martin (1995)). In practice, research has been driven by data availability and the development of methods comparing the relative explanatory power among the many potential determinants (see Sala-i-Martin et al. (2003)). The following analysis draws a number of parallels with IMF (2004), which analyzes 17 OECD economies. However, it differs with respect to the specificity of the European focus and the nature of the data.

14. In the context of the EU, the key “priors” involve interaction of pan-European, national, and subnational causes of growth. This distinction would be consistent with the stylized facts on the EU’s current economic and policy environment, such as: (i) a drive toward economic policy convergence, whether with respect to regulations or lower barriers to the circulation of goods, services, and factors of production; (ii) the remaining scope for national policies and the less-than-perfect liberalization of cross-border flows and regulations, and (iii) developments influencing within-country variation in regional growth. In line with this three-fold distinction, per capita growth (expressed in the standard log-form) in a given region i in country k at time t could be expressed as a function of contemporaneous and lagged region-specific, national, and “global” factors:

Dyit=F(Ri,(t,t−1),Nk(i)(t,t−1),G(t,t−1))(2)

where D is the usual difference operator, R stands for region-specific factors (including convergence); N denotes nation-wide variables, and G would be a collection of common/global factors.

15. Among these three levels of analysis, national factors merit greater attention. First, despite ongoing efforts and progress in enforcing competition, regulation, and macroeconomic management at the EU-level (Blanchard (2004)), national macroeconomic and structural policies continue to play a central role in economic outcomes (see Sapir (2005)). Second, high-quality nationwide data are universally available, while the range of relevant regional data has been much more limited at the EU-wide level.10 As to the global factors, they may explain performance of European countries compared to other parts of the world, but, generally, would not explain variation within Europe without an explicit consideration of their differential impact on different regions.

16. Cross-country literature on advanced economies suggests multiple nationwide structural policy dimensions for augmenting the growth regressions. IMF (2004) singled out five such areas for OECD countries: (i) product markets; (ii) trade reform; (iii) labor markets; (iv) fiscal reform and (v) financial reform. In addition to these, the cross-country regressions include controls such as the private investment to GDP ratio, terms of trade and population growth, measures of the stock of human capital, lagged GDP per capita level, and financial development.

17. In the context the EU, a few specific structural rigidities have been considered particularly important.

  • A high tax burden, which is pointed to by many authors (Prescott (2004), Neil and Kirkegaard (2004)) as a major factor affecting incentives for labor and investment.

  • Lack of regulatory flexibility in labor markets and high unionization, which could constrain labor supply (see Alesina et al. (2005)).

  • Regulation of product markets, which, by limiting competition, can affect productivity and medium-term growth performance (Blanchard (2004)).

  • Poor business environment, including rigid bureaucratic and legal processes that discourage investment.

  • Lack of financial sector development, which, in line with contributions by Rajan and Zingales (1998), may have limited entrepreneurial opportunities, especially given the few options of arms-length financing in continental Europe.

These rigidities, while mostly (but not always, see Appendix) linked to nationwide policies, would clearly affect regional performance in the relevant countries.

18. The empirical strategy is thus anchored to the interaction of regional convergence and the above nationwide factors in explaining (regional) growth. In particular, equation (2) would be tested in the form of:

(3)Dyi,t=c+αyi,t−1+{βPMRk,t−1+γEPLk,t−1+Taxk,t+δDPMRk,t+ϕDEPLk,t}+λX(t)+μi,t

The first term controls for convergence at the regional level. The expression in brackets seeks to capture the key country-level factors both in terms of their lagged levels and the rates of change. In particular, the following measures have been tested: (i) PMR is a comprehensive measure of product market rigidities, which is available from OECD as an index for two years: 1998 and 2003; (ii) EPL denotes labor market rigidities, in the form of employment protection legislation (EPL), which is also available from OECD in index form for the end-of 1990s and for 2003; and (iii) Tax reflects the extent of the tax burden, with various alternative measures standing for the total and its subcomponents (taxation of capital and labor), which are available from Martinez-Mongay (2003).11 X denotes other potential (region-specific, nationwide, or global) control variables and interaction terms, including those referring to the business environment and financial development, which may or may not be time-specific (and will be reported below where relevant).

19. The focus on the closely-knit EU economies, and the use of recent higher-quality data, permit a reasonably comprehensive but still parsimonious specification. In particular, a number of customary cross-country controls may be omitted for the EU countries, given their limited variation across regions (due to, for example, the common currency and the setting of some regulations on an EU-wide basis). In addition, the OECD’s economy-wide PMR index captures not only product market regulation per se, but also rules affecting domestic and external trade; this was not the case for the narrower such indices that were previously available. Thus, the data underlying equation (3) permit coverage of structural factors broadly comparable to that in IMF (2004), despite using a smaller number of explanatory variables (see Appendix).

20. Although the period and method of analysis is constrained by data availability, pronounced data variation accentuates the key relationships. OECD data on labor and product market regulation are only available for late 1990s and 2003, thus (largely) overlapping with NUTS-2 data for 1998–2002. The other variables were chosen to match this latter period, in a cross-section regression (see Appendix for the definition and sources of the variables). The data are for the EU-14 countries, as Luxembourg was excluded due to the lack of the data on product market regulation in 1998. The key structural data over this period exhibited marked variation, as the initial level of the indexes was generally inversely correlated with the subsequent change, with the most regulated countries (notably Italy) experiencing greatest improvement, especially in terms of the PMR index (see Figure 2).

Figure 2.
Figure 2.

EU-14: Product and Labor Market Indicators, 1998-2003.

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

Source: OECD
Results

21. Cross-section regressions suggest strong negative effects on growth of the tax burden (notably on labor) and of structural rigidities in product markets over 1998–2002. The main results are presented in Table 5; they are largely in line with the related literature, and can be summarized as follows:

  • In contrast to the above regressions with country dummies, regional convergence generally holds with additional controls, although the coefficient is somewhat smaller and less significant (but still significant at the 5 percent level) than in absolute convergence regressions. It is thus the case that differences in national fiscal and structural policies could be partly driving the observed unconditional convergence process.

  • The effect of the nation-wide tax burden on economic growth is negative, large, and highly statistically significant. While the impact of the general tax burden indicator (tax/GDP) is strongly negative, there is substantial difference in the effects of its “subcomponents,” with labor taxes having a very significant negative association with growth, and capital taxes having a counterintuitive effect. This result is interesting in the context of the recent debate between Alesina and others (2005) and Prescott (2004) on the relevance of the tax rate as an explanatory factor for labor supply in Europe. It would seem to reinforce Prescott’s claim about the adverse effects of the high labor tax burden, especially as the labor market protection emphasized instead by Alesina and others is controlled for in these regressions. However, the dependent variable is not labor supply but growth, and the tax rate could feasibly affect it through channels other than labor supply (including through the correlation of the labor tax burden with other fiscal policy indicators). Still, the results are suggestive of relatively large “macroelasticities” of the labor tax rates for the 1999–2002 period.

  • National product market regulation is negatively (and significantly) related to regional growth (as an average over a 4-year horizon), both in terms of the lagged impact of the index of regulation and the contemporaneous effect of the change in the index. This effect is present both with and without the tax burden variable, although it is weaker in some specifications. The qualitative result compares to that of the IMF (2004), which establishes that product market reform would have a moderately positive contemporaneous effect that becomes slightly negative over a 3-year horizon and turns sharply positive thereafter. However, the broader specification of the new PMR index also includes trade reform indicators, which in the IMF (2004) specification is assessed as having a separate positive effect on growth. Adjusting for this difference, the results seem qualitatively consistent.

  • The degree of labor market regulation is, however, “positive” for growth, although the economic significance is small. This may not be surprising in the short term, however, as more labor market flexibility can lead to wage moderation and a short-term contraction in demand (IMF (2004)). The statistical significance in the level of the index is somewhat surprising and appears to contrast with IMF (2004), which found that over a longer-term horizon the effect of increased labor market flexibility would be positive for growth. Some features of the current specification may, however, affect this finding: (i) the effect of labor market reform in the period prior to the initial measurement of the index conflicts with the interpretation of the index as embedding only long-term factors; and (ii) the index may be too narrow as a summary measure of the degree of labor market flexibility. In the latter case, as argued by Alesina and others (2005), the residual labor market rigidity could be correlated with the tax rate.12

The proposed baseline regression is reported in the fourth column of Table 5; it includes the labor tax burden, and the lagged level and contemporaneous change in the PMR index, but excludes the EPL index, given the instability in some of the above results.

Table 5.

Europe-wide Regional Growth and National Policy Environments, EU-14 (excluding Luxembourg), Basic Cross-section Regression, Dependent variable PPP real Change per Capita in 1999–2002, log-form

article image
Note: regressions include a constant (not reported), t-statistics in parentheses. * denotes significance at 5 percent and **at 1 percent

Evaluated with robust standard errors. Instruments are selected among the components of the PMR index based on their correlation properties that would maximize their validity.

22. Controlling for potential methodological problems does not alter the thrust of the results. There could be four principal caveats to this analysis: (i) possible endogeneity or simultaneity biases; (ii) multicollinearity in some regressors; (iii) a potential dominance of short-term cyclical factors in driving the results; and (iv) distortions in combining national with regional data. These problems are discussed in the Appendix (including by instrumenting for potential reverse-causality for some variables), and do not appear to have affected the results, although some issues remain open.

23. An analysis of components of the PMR index indicates that “economic,” as opposed to “administrative,” liberalization had a greater short-term association with growth. Table 6 shows how the values and significance of the various subcomponents of the index performed as alternative explanatory variables for growth (in lieu of the aggregate index). Of the two main subcomponents, economic regulation was statistically significant both for the level and the rate of change, while administrative regulation was only borderline significant for the level and not significant in terms of change in the index. Among the 16 primary subcomponents of the PMR index, the expected effects appeared strong and significant with respect to: (i) barriers to trade, investment, and entrepreneurship; (ii) license and permit system; (iii) size of the public enterprise sector; and (iv) regulatory barriers. On the other hand, such factors as (i) simplification of rules and (ii) absence of command and control regulation had a counterintuitive inverse association with growth over this period. This, however, may reflect the either the possible J-curve effect or methodological problems, especially given the partial nature of these explanatory variables.

Table 6.

Subcomponents of the Product Market Regulation Index in the Convergence Regressions, eu-14. Dependent Variable: PPP Real Growth per Capita in 1999–2002, log-form

article image
Note: Regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent.

The coefficients that passed robustness checks are marked in bold.

24. Measures of financial sector development do not play a stable role in these regressions, possibly reflecting data issues. As per Table 7, beginning-of-period arms-length financial development indicators, such as capitalization of stock market/GDP ratio in 1975–1995 (or in 1998), did not add any explanatory power to the baseline regression. The intermediary credit-to-GDP ratio had a positive, but small and statistically insignificant, relationship with growth. This may, however, reflect the fact that financial depth may vary not so much between countries as between regions in the same country, and the latter is not incorporated in the regressions. In addition, some of the elements of the PMR index may already capture aspects that are highly correlated with financial sector development/reform.

Table 7.

Role of Financial and Institutional Development measures. Dependent Variable PPP: Real Growth per Capita in 1999–2002.

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent.

25. Alternative nationwide institutional factors - the length of legal procedures and rate of recovery of firm assets in bankruptcies - add explanatory power in line with intuition (Table 7). The inclusion of the length of civil legal procedures (available only for the year 1996, which roughly approximates the beginning-of-period sample) predictably has a negative association with growth. The asset recovery index (from the World Bank’s “Doing Business” dataset) also has the expected positive sign with respect to growth, and is highly significant. However, its inclusion in the regression substantially weakens the significance of the PMR index (though the rate of change of the PMR would still remain significant).13

26. Measures of competitive pressure from China appeared to have little explanatory power, although the limited number of observations likely affected the results. The key challenge involves finding suitable aggregate measures of such competition for the EU-14. There is a growing literature on measures of structural competitive pressure on exports, notably De Nardis and Trau (1999) and Monti (2005), who provide methodological and empirical contributions with particular reference to Italy, but no crosscountry “aggregate” measures. A simple measure could be given by the extent of overlap in Balassa sectoral specialization indices of the EU-5 countries’ manufacturing exports with those of China (see Appendix). On this basis, competitive pressures on Italy appear indeed substantially higher than on other large EU countries (see Figure 3).14 As can be seen from Table 8, the additional effect of this measure on EU-4/5 growth has not been robust. Still, the result hinges on the very few country observations and the full baseline specification of equation (3) could not be tested because of the lack of cross-country observations. Moreover, China is only one, albeit very important, element of competition from the emerging markets.

Figure 3.
Figure 3.

EU-5: Pairwise Products of Sectoral Symmetric Balassa Indices of EU-5 countries with China, 1997

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

Source: SVIMEZ (see Appendix).
Table 8.

Europe-wide Regional Growth and National Policy Environments, EU-4/EU-5, Cross-section Regression, Dependent variable PPP Real Growth per Capita in 1999–2002.

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent

27. The role of country circumstances is illustrated by the analysis of Italy and Germany dummies. These two countries are useful to compare, given that they have both recorded slow growth, although their experiences have differed in terms of the role of competitiveness. Table 9 shows that the inclusion of the Italy dummy, at the margin, significantly weakens the effect of the beginning-of-period product market regulation index on growth, suggesting that the latter plays an important role in the relative performance of Italy’s regions (as a group). On the other hand, the inclusion of the Germany dummy appears to weaken appreciably the effect of the tax burden on labor, thus indicating that this burden -conditioned on the model - is a significant drag on growth in the German regions. Still, even with the inclusion of the dummy, the coefficient appears amply significant, suggesting that the high level of labor taxes has been associated with lower growth in other European countries as well. Interestingly, France or Spain dummies (not shown) do not noticeably affect the magnitude of these coefficients at the margin.

Table 9.

The Role of Italy and Germany Country Dummies, Dependent Variable PPP Real Growth per Capita in 1999–2002, EU-14

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent

28. A number of further variables were tested but did not add much explanatory power to the regressions. In particular, this regarded data on investment/GDP ratios; Foreign Direct Investment; R&D expenditures; human capital, measures of infrastructure endowment (intensity of railway and highway networks); and various macroeconomic indicators (beginning-of-period and average inflation, budget deficits, employment rate). These could have been insignificant for different reasons, however, including (i) relative variations in the data, which may well be more pronounced between individual regions than between nations,15 particularly for FDI and infrastructure; (ii) quality/comprehensiveness of the measures, for example infrastructure endowments are difficult to proxy in a comprehensive way; and (iii) the timeline effect of growth of some of the variables, particularly those related to the macroeconomic environment.

C. Italy’s medium-term regional growth

29. An EU-wide focus may neglect important country-specific aspects. In the above analysis, data limitations could have led to an omission of potentially relevant within-country, region-specific factors. In addition, the methodology for deriving some of the above country-wide variables may not be fully uniform. Alternative, and often more detailed, sets of data are available for national economies. For example, these would permit a characterization of various external and sectoral patterns, information on which is lacking at the NUTS-2 level of disaggregation.16

30. Italy is an important case study of growth dynamics, not least because of the interplay of European, nationwide, and local factors in its overall growth record. A regional analysis of Italy may be particularly insightful given the wide territorial divergences. The coefficient of variation of income across regions, at about 25 percent, is higher than in most large European economies. There has been substantial research into many aspects of the deep differences in Italy’s regional economic performance, which has been dominated by North-South issues (see Vamvakidis (2003) and references therein). In particular, variation in the size and nature of investment, institutional quality, and human capital has been traced to unfavorable economic outcomes in many Southern regions. Excessively rigid centralized wage bargaining (see Vamvakidis (2002)) may also have contributed to regional growth variations by inhibiting labor market clearing.

31. This section focuses on Italy’s recent medium-term performance. In line with the previous section’s framework, within-country regional growth could be assumed to depend on (i) convergence; (ii) policy variables, (iii) institutional quality, and (iv) other factors specific to the particular context, for example shocks that affect the regions differentially. However, consistent with available studies on regional growth, many of the typical policy and institutional variables do not need to be included, since they are often identical (or nearly so) for all regions. Regarding structural policies, most of these are in the purview of European and national authorities. To the extent these are influenced locally, no consistent data are available, in any case.

32. A key issue widely debated in regard to Italy’s growth in the past decade is its sectoral trade specialization, which also exhibits substantial regional variation. As per Figure 3, Italy’s manufactured exports have on the whole been more similar to those of emerging markets than those of other advanced economies. This “similarity” has been highest for leather/footwear and textile products. As follows from Figure 4, specialization in production and exports of these two sectors exhibited pronounced variation across regions. In particular, textiles as a share of regional manufacturing value-added varied from a minimum of 2 percent in Liguria to some 20 percent in Tuscany. Leather and footwear, while generally very small as a share of manufacturing value-added, was quite important in Marche and, again, Tuscany. The picture is similar in terms of regional export specialization. It would thus be interesting to gauge whether these differences in the “legacy” of sectoral specialization can be robustly linked to economic outcomes.

Figure 4.
Figure 4.

Italy, Asia-prone Production in Regional Manufacturing and Exports, 2001–2004.

(in percent)

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

Source: ISTAT and Bank of Italy.

Other, more commonly accepted, growth determinants at the regional level regard financial development and investment, although their impact would be uncertain over the medium term. Financial development varies quite substantially across Italy’s regions, and it has been linked to growth outcomes (Guiso and others (2002)), although the crosscountry regressions reported earlier did not find it to be significant. At the same time, causality issues are particularly tricky with respect to the interaction of financial markets and growth, given the well-known hypothesis that the former may anticipate the latter (Rajan and Zingales (1998)). Regarding regional investment, Vamvakidis (2003) found some weak positive long-term effects of the infrastructure investment/GDP ratio. However, even over a long time frame, the effect of the non-infrastructure investment/GDP ratio was insignificant.

33. The estimated equation for real per capita growth thus draws on cross-country and Italy-specific literature, and would rely on panel data for greater testing power. Thus, the key ex-ante variables, in addition to convergence, would be those denoting lagged and/or contemporaneous trade and sectoral structure and financial development. The estimated model would be as follows:

Dyi,t=c+αyi,t−1+structurei,t,t−1+findevi,t,t−1+(inv/Y)i,t,t-1+otheri,t+μi,t(4)⁢

where structure would denote a variable characterizing Italy’s export or production specialization, findev incorporates available regional financial sector variables, the ratio of regional investment to GDP, and “other” would include additional potentially important variables in explaining regional growth. We would proceed by sequentially augmenting equation (3), which would allow to pay particular attention to “structure,” while considering other variables as controls. Of course, the caveats related to the sample length discussed in the cross-country section continue to apply.

34. The absolute convergence among Italian regions found in the NUTS-2 level data also holds in terms of per capita GDP. Using the data for Italy’s 20 regions produced by ISTAT for 1995–2002 yields a coefficient of 1.7 percent for 1995–2002 for convergence in terms of real per capita GDP (comparable to 1.4 percent as found above in NUTS-2 PPP-adjusted data for Italy). Figure 5 shows that convergence held both during 1997–2000 and 2001–04. There were, however, important differences between these two subperiods, in that in the first period all regions were growing, while in the second convergence occurred against the backdrop of output declines in a subset of regions. This shift was also accompanied by a steeper slope in the convergence line in 2001–04, which suggests hat the variation among regions may offer a relevant dimension for tracking the slowing of Italy’s aggregate growth. In particular, one may check whether the absolute declines in output in some of Italy’s regions could have been related to its specialization in products that are subject to increasing competition from emerging market economies.

Figure 5.

Italy: Real GDP per Capita Convergence.

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

Source: ISTAT

35. Superficial evidence on the interaction of sectoral and regional patterns suggests that the regional trade specialization in manufacturing did not have a strong association with regional growth. Figure 6 augments the unconditional convergence graph for the cross-section of annual averages over four years (2001–04), by linking it to the extent of manufacturing export specialization in the key “Asia-competing” sectors: textiles and leather/footwear. The extent of regional specialization, given by the export specialization (Balassa) index at the beginning of the period, is denoted by the size of the bubbles. There does not appear to be a clear pattern with respect to textiles, as the bubbles are distributed more-or-less randomly along the convergence line. There appears to be a somewhat greater link with leather and footwear, whereby slower-growing regions in the lower-right corner tend to have a somewhat higher export specialization in the sector. However, the weight of the latter sector is quite small in the overwhelming majority of regions. In any case, a more formal analysis is needed to check this link under various controls.

Figure 6.
Figure 6.

Italy: Regional Convergence and Asia-prone Sectoral Shares, 2001–04.

(bubbles denote Balassa sectoral specialization indices)

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

Source: ISTAT and Bank of Italy

36. Regression analysis also suggests a limited role of export specialization in the regional variation of output over the last few years. Table 10 shows the results for measures of specialization in exports for 2001–04. In simple pooled OLS regressions, the coefficient on export specialization in textiles was neither economically nor statistically significant. While the coefficient on leather and footwear was just statistically significant at the 5 percent level, likely because of a sharper variation of the share of these products across regions, its economic significance was very small. Exploiting the time variation in the data through a fixed-effects regression renders the both sectoral specialization coefficients statistically insignificant, with the export specialization in textiles actually having a counterintuitive positive sign. The GLS “between/within” specification yields negative signs, but the coefficients are statistically insignificant for both sectors. Among the control variables in the 2001–04 data, convergence is highly statistically significant, but financial sector measures are not (share of financial intermediation in GDP (not shown), lagged and contemporaneous credit growth).

Table 10.

Italian Regions: Dependent Variable: Change in Log Real per Capita GDP, panel data, 2001–04.

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent

Evaluated with robust standard errors.

Logs of regional Balassa export specialization indices.

37. The weak significance of the measured specialization does not change with the extended timeframe and addition of other variables. Table 11 presents the results for the 1995–2003 period. In particular, the (lagged or contemporaneous) share of textiles and leather and footwear in value added is not significantly associated with regional growth. A “combined” share of all such sectors, including rubber, did not change this result. The role of other variables (controls) is roughly in line with the literature on regional growth in Italy (Vamvakidis (2003)), with a significant influence of convergence, marginal significance of the “South” dummy (negative), and insignificant role of the share of total investment in regional GDP. Time dummies (not shown) generally continue to exhibit a downward trend over time, indicating that the substantial worsening of performance at the end of the period remains fundamentally unexplained. The results for the specialization are broadly the same under different econometric specifications, including those that take into account potential endogeneity of the variables (General Method of Moments estimator).17

Table 11.

Italian Regions: Dependent Variable: Change in Log Real per Capita GDP, Panel Data, 1996–2003

article image
Note: regressions include a constant (not reported), t-statistics in parentheses.

denotes significance at 5 percent and

at 1 percent. All regressions include time dummies.

Evaluated in first differences, Sargan test: 3.926.

Also evaluated with logs; results were essentially similar.

38. A number of contributions suggest substantial role of unfavorable specialization, in exports and, ultimately, country-wide growth outcomes. In particular, Bugamelli and Rosalia (2004) offer evidence that the sectors associated with competition from China experienced lower growth in output and exports than other sectors. Similarly, Chapter II of this set of papers cites other work that seem to arrive at the same conclusion with respect to external market shares. However, from the analysis above, it appears that these differences have not affected the relative growth of Italian regions.

39. The low role of sectoral export specialization in regional growth found here need not, however, conflict with the above literature. First, the unfavorable effects of exports on growth could be in practice dampened due to offsetting effects of domestic demand and reallocation of resources between regions within Italy. Second, the simple sectoral disaggregation used here may be too coarse for pinning down the effects of specialization, and some authors have studied this at a more disaggregated level.18 Third, the various controls typical for the growth literature, notably convergence, may not work as effectively in filtering short-term growth regressions, as in the longer-term regressions, thus possibly imparting a bias to some results. Still, convergence itself has been surprisingly robust, stable and unaffected by the “short-term bias” compared to the sectoral specialization, which warrants further research on the issue.

D. Conclusion

40. This chapter offers an analysis of the recent medium-term growth experience in the EU and Italy. It concludes that: (i) regional EU-wide convergence was observed in 1995–2002, though largely between, rather than within, countries, (though within Italy, convergence was significant over the past decade); (ii) greater national product market regulation and a higher tax burden were associated with lower growth in 1999–2002; and (iii) unfavorable sectoral “legacies” may have played some role in explaining regional growth variations among some EU countries and in Italy in particular over the last decade, but their impact was typically small.

41. This research complements cross-country empirical literature on growth and structural reforms in the EU. First, it confirms and extends evidence in favor of ongoing Europe-wide convergence (Vamvakidis (2003)) by finding it in the larger EU-25 sample, and even within fairly short (3-year) intervals. Second, liberalizing product market regulation and lowering the tax burden are found to be important for jumpstarting Europe’s growth, in line with evidence by Nicoletti and Scarpetta (2003) derived on the basis of firm-level crosscountry data. Regional data permit an interpretation of this conclusion in broader terms, using a “whole-economy” PMR measure. Third, the chapter complements vast cross-country literature on the impact of institutions on growth (Acemoglu and others (2004)), in terms of its applicability to advanced economies, were the relevance of existing global classifications (i.e., the World Bank’s “Doing business” indicators, Transparency international, World Economic Forum, etc.) for empirical research has sometimes been questioned.

42. With respect to Italy’s performance, European and country-specific data point to the primacy of structural and fiscal policies. On the basis of Europe-wide analysis, the main priorities for Italy would be to pursue liberalizing structural reforms, especially product market deregulation. To the extent additional fiscal consolidation permits durable reductions in the tax burden, cross-country results suggest that labor tax cuts could be particularly effective. Regarding structural policies, however, it is somewhat less encouraging that Italy’s improvement in the OECD PMR index between 1998 and 2003 was hardly reflected in its growth record. One partial explanation is that the index, while being an important objective measure of the formal regulations, does not capture “informal” factors affecting their implementation. If so, the role of de-facto policy improvements may be even greater than the regression results suggest.

43. Further research may be needed to confirm or refine some conclusions, especially as data availability and quality improve. The period of analysis is quite short, and the association of growth with structural variables, particularly in terms of the rate of change of the latter, would need to be investigated further with an extended time series of economy-wide structural data. A more representative dataset would also enhance the power of the panel regressions, wherein cyclical factors are so far controlled very imperfectly by time dummies. More detailed data, at the Europe-wide, national, and regional levels, would permit more sophisticated econometric tests, exploring the time series dimension with greater precision to better capture the evolving institutional and policy environment, including at the subnational level and more specific structural reforms. Even against these imperfections in the observed period, both the integration process and the remaining rigidities have been strong enough to be measurably associated with growth even over a short-to-medium-term span.

A. Data description.

Europe-wide data:

The NUTS-2-level data are available for 254 (or in some samples 258, including 4 French territories) European regions of the EU-25, on Eurostat’s online Regio database http://epp.eurostat.cec.eu.int/portal/page?_pageid=1090,30070682,1090_33076576&_dad=p ortal&_schema=PORTAL. These specific variables include: (i) PPP-adjusted GDP for 1995–2002; (ii) employment; (iii) population; (iv) education; (v) spending on research and development, and (vi) sectoral composition of business activities. Other potential control variables from the NUTS-2 disaggregation are also available, but have not been used, partly because of the large data gaps across time or sectors.

Several tax burden indicators are from Martinez-Mongay (2003) for EU-15 countries through 2002, including: (i) average effective tax rate on labor income, which is the ratio of the sum of non-wage labor costs plus the personal income tax revenue attributed to labor income; and (ii) average effective tax rate on capital income, which is the sum personal income from capital, taxes on corporate income, and property taxes. The tax/GDP ratio was taken from Eurostat.

Product Market Regulation (PMR) index is an objective measure of the formal rules governing product market regulation across the whole economy, collected over 16 primary measures (low-level indicators) in the three broad areas of (i) state control; (ii) barriers to entrepreneurship; and (iii) (external) barriers to trade and investment for most OECD countries. A consistent aggregation of the different components is achieved through the method of principal components. The data were first compiled for the year 1998 and were later updated for 2003, together with the revision of 1998 data. (See Conway and others (2005) for further details). Among the EU-15, data for Luxembourg are not available for 1998, and thus most regressions exclude Luxembourg.

The Employment Protection Legislation (EPL) index refers to restrictiveness of (i) protection of regular workers against dismissal; (ii) specific requirements for collective dismissals; and (iii) regulations of temporary forms of employment. Some factors, including the role of contractual provisions and judicial practices, do not get reflected in the indicator. The data are available for late 1990s and 2003 for OECD countries. For more information see OECD (2004).

Measures of financial development comprise (i) stock market capitalization ratio to GDP in 1975–95 and for 1998 (source: Beck et al. (2001)); and (ii) intermediary credit to GDP ratio (source: Beck et al. (2001)). .

Length of civil legal procedures is measured by data on average duration of the three degrees of judicial process, expressed in months. These are available for all EU countries for 1996 (see European Commission (1998)).

Rate of asset recovery (in terms of cents on the dollar) is available from the World Bank’s “Doing business” indicators database, for the year 2004. Given that no corresponding data exist for the previous period, the underlying assumption is that the cross-country variability with respect to this statistic had not been changing much between 1998 and 2004.

A measure of the similarity of exports and export specialization in manufacturing. The data for this calculation have been provided by Svimez for 1997 and 2002 (on the basis of a 3-digit-disaggregated export commodity classification (ATECO)), whereby Balassa indices have been calculated for 17 large advanced and emerging countries, only 5 of which are members of the EU (Italy, Spain, France, Germany and the UK - EU5)).

Balassa indices. The common Balassa index of revealed comparative advantage (RCA) is defined as follows:

RCAij=Xij/∑jXij∑iXij/∑i∑jXij

where the ratio in the numerator is the share of country/region j in sector i world manufacturing exports (on the basis of the Global Trade indicators (GTI) database, which covers around 80 percent of world trade), while the ratio in the denominator represents the same share for total exports. The index varies between zero and infinity, with values greater than unity denoting the presence of “positive” specialization.

The symmetric Balassa index is a monotonic transformation of the RCA index set to vary between -1 and 1, with values greater than zero representing “positive” specialization:

RCASij=RCAij−1RCAij+1

The “China overlap” in the Balassa indices (a variable used in EU-4 and EU-5 cross-country regressions) is computed as a weighted (by manufacturing trade shares of a given EU country) sum of pairwise products of sectoral symmetric Balassa indices (provided both are positive), of China and the relevant EU country respectively.

Italy-specific data:

These are available from ISTAT, Italy’s statistical authority (www.istat.it), including: (i) regional GDP per capita in constant and current prices; (ii) gross fixed capital formation; (iii) value added in industry and in individual industrial sectors; (iv) employment; and (v) wages.

Bank of Italy (2001–04) was used for regional financial sector data and regional export specialization indices.

B. Methodological remarks.

The basic specification for Europe-wide data is the cross-section growth regression of Barro and Sala-i-Martin (1995), given that there is no extended time series for whole-economy structural data. While these simple regressions do not permit testing dynamic effects of reforms in line with the recent panel data-based research, they are still extensively used (see Sala-i-Martin and others (2003)), and offer a preferred initial test for cross-country relationships. With respect to the effect of structural reform on growth, using only crosscountry data in a cross-section regression would limit the power of the tests (which would be based on 14 observations). Subnational data increase the power of these cross-section tests, but the specifications have to be parsimonious given the limited number of countries and the likelihood of specification problems once the number of country-wide explanatory variables becomes very large in a single regression. The regressions would however have some potential methodological problems and caveats.

Multicollinearity. Some regressors appear to have potential for being collinear. However, the bilateral correlation coefficients (not reported) are generally not so high as to indicate the multicollinearity problem, with the possible exception of the levels of the PMR and EPL indices, which exhibit bilateral correlation of around 90 percent. The latter variable, however, is not included in the baseline specification. In any case, none of the regressions (including those with the EPL measure) appear to exhibit the common multicollinearity symptoms of high R-squared accompanied by high standard errors.

Reverse causality. Instrumental variable regressions (estimated by two-stage least squares) have been used to check for potential endogeneity between the concurrent change in the product market regulation index and growth; these regressions are reported in the right-hand part of Table 5 and do not affect the signs and significance of the variables (but do affect the magnitude of some elasticities). In practice, reverse causality from growth to the explanatory variables is very unlikely in these regressions because of the inverse relationship between the level and the change in the PMR index over the observed period (Figure 2. (In case of significant reverse causality these are more likely to be positively correlated). In addition, a priori theoretical link from reforms to growth is ambiguous, and possibly non-linear. (Reforms have been considered easier to implement during spells of good growth, but also in times of crises).

Another possibility is the reverse causality between GDP growth and the tax burden, with the latter being partly driven up by slow growth. In this respect, one may note that a substantial inertia in the levels of cross-country tax burdens makes the reverse causality over the medium-term growth horizon less likely, except in the event of substantial long-term prior inertia in differences in cross-country growth rates. In any case, instrumenting for the link did not change the qualitative signs of the coefficients.

“Shortness” of time horizon. The basic cross-section regression for the EU covers a 4-year time span of 1999–2002, while other EU and Italy-specific regressions have been estimated over the horizon of 4–8 years. These periods are short, but not unprecedented in the analysis of growth, especially if annual averages are used for EU data. IMF (2004) employed 3-year intervals to smooth out “short-term” cyclical fluctuations for GDP. Furthermore, the particular horizon chosen limits within-cycle biases. In this respect, the 1999–2002 period in the EU, and the 1996–2004 period in Italy (and the estimated subsamples) as annual averages have captured, roughly in equal measure, periods of cyclical strength and weakness. Also, both conditional and unconditional convergence coefficients for the augmented regional sample have been relatively stable and consistent with “standard” values over these medium-term horizons.

More generally, Jones and Olken (2005) criticize the prevailing focus on the literature on long-term growth experiences, as these appeared to be a summation of very different sequences of medium-term episodes, which, as they argued, have to be analyzed separately. (In addition, long-term growth regressions are likewise susceptible to biases arising from different temporal profile of beginning-of-period controls). The medium-term cross-section analysis allows a focus on the characteristics of the key period, incorporating its unique features. It is thus encouraging that the qualitative role of most (structural and other) explanatory variables has been in line with intuition. Finally, the expanded number of cross-sectional observations partly compensates for the shortness of the horizon.

Particular combination of country-level versus region-specific data. The chosen mix of country-level indicators and region-specific variables may not be fully optimal for EU-wide regressions. Intuitively, not all country-wide variables would be appropriate for explaining regional growth performance within the EU. In particular, within-country variation should ideally dwarf cross-country variation (otherwise there would be a need for the corresponding region-specific information). In this respect, macroeconomic variables, the tax burden, and structural reforms would largely fit these criteria, since most of their aspects are (so far) determined at the national level in the EU countries, although some of this may change as devolution of power to the regions progresses in some countries. Financial development may however differ among regions in one country quite substantially.

Two efforts have been made to further control for this problem. First, other available data on key country-level and region-specific (NUTS-2) indicators have been tested as additional controls, but generally have not added explanatory power. Second, the configuration of country-level factors is consistent with that of other studies, in particular IMF (2004). Figure 7 shows that there is a broad similarity in the dimensions of structural reform covered by the two studies. The main differences of the present study are (i) narrower treatment of the tax reform and labor market indicators - prompted by the desire to focus on key single aspects of specific dimensions as opposed to “unweighted sectoral averages;” and (ii) omission of “financial reform” measures, whose effects turned out fairly weak in the IMF (2004) study, and given the presence of controls for financial development in some specifications.

Figure 7.
Figure 7.

Comparison of coverage of structural reforms with the IMF (2004) study.

Citation: IMF Staff Country Reports 2006, 059; 10.5089/9781451819885.002.A003

References

  • Acemoglu D., Johnson S., and J. Robinson (2004) “Institutions as the Fundamental Cause of Long-Run Growth,” NBER Working paper, 10481, May.

    • Search Google Scholar
    • Export Citation

  • Alesina A., Glaeser E., and Sacerdote B. (2005) “Work and Leisure in the US and Europe: Why So Different?,” available at http://post.economics.harvard.edu/faculty/alesina/papers/work_leisure.pdf

    • Search Google Scholar
    • Export Citation

  • Bank of Italy (2004, 2005) “Relazione del Governatore,” Rome, May.

  • Bank of Italy (2001 –04) “Summary of the Reports on economic developments in the Italian regions,” Banca’d’Italia, Rome.

  • Barro R. and X. Sala-i-Martin (1995) “Economic Growth,” McGraw Hill, N.Y.

  • Beck T., Demirgüç-Kunt A and R. Levine (2001) “Legal Theories of Financial Development,” Oxford Review of Economic Policy, Vol. 17, No 4, 483 –501.

    • Search Google Scholar
    • Export Citation

  • Blanchard O. (2004) “The Economic Future of Europe,” NBER Working paper, No 10310, March.

  • Boldrin and F. Canova (2001) “Europe’s Regions: Income Disparities and Regional Policies,” Economic Policy, No 32.

  • Bugamelli M. and A. Rosalia (2004) “Produttivita’ e concorrenza estera,” Banca d’Italia, Servizio studi, mimeo.

  • Canova F. and A. Marcet (1995) “The poor stay poor: nonconvergence across countries and regions,” Discussion paper No 1265, CEPR, London, November.

    • Search Google Scholar
    • Export Citation

  • Cerra V. and S. Chaman Saxena (2005) “Growth Dynamics: the Myth of Economic Recovery,” IMF Working paper, 05/147, July.

  • Conway P., Janod V., and G. Nicoletti (2005) “Product Market Regulation in OECD Countries: 1998 to 2003,” OECD/WKP (2005)6.

  • De Nardis S. and F. Traù (1999) “Specializzazione settoriale e qualità dei prodotti: misure della pressione competitiva sull’industria italiana”, in Rivista Italiana degli Economisti, n. 2, 1999, pp. 177 –212.

    • Search Google Scholar
    • Export Citation

  • Decressin J. (2000) “Puzzling Out Italy’s Growth Performance,” IMF Staff Country Report 00/82, July, 6–27.

  • European Commission (1998) “The Cost of Legal Obstacles to the Disadvantage of Consumers in the Single Market” report Edited by Von Freyhold, Vial and Partner Consultants, DG XXIV, Brussels.

    • Search Google Scholar
    • Export Citation

  • Guiso L., Sapienza P. and L. Zingales “Does Local Financial Development Matter?,” NBER Working paper, No. 8923, May.

  • International Monetary Fund (2004), Fostering Structural Reforms in Industrial Countries, Chapter 3 in World Economic Outlook, April.

  • Jones B. and B. Olken (2005) “The anatomy of start-stop growth” NBER Working paper, 11528, August.

  • Krugman P. (1991) “Increasing returns and economic geography,” Journal of Political Economy, 99.

  • Levine R., Loyaza N., and T. Beck (2000) “Financial Intermediation and Growth: Causality and Causes,” Journal of Monetary Economics, 46 (1), 31 –77.

    • Search Google Scholar
    • Export Citation

  • Martinez-Mongay C. (2003) “Labor Taxation in the European Union. Convergence. Competition. Insurance?” in pp 31 –68 in “Tax Policy” Banca d’Italia, Research Department, proceeding of the Public Finance Workshop, 2003.

    • Search Google Scholar
    • Export Citation

  • Monti P. (2005) “Caratteristiche e mutamenti della specializazzione delle esportazioni italiane,” Temi di Discussione del Servizio Studi, Banca d’Italia, No 559, September.

    • Search Google Scholar
    • Export Citation

  • Neil M. N. and J. F. Kirkegaard (2004), Transforming the European Economy, (Washington: Institutte of International Economics).

  • OECD (2004) “Employment Protection Regulation and Labor Market Performance,” Chapter 2, in OECD Employment Outlook, OECD, Paris.

  • Prescott E. (2004) “Why do Americans work so much more than the Europeans?” Federal Reserve Bank of Minneapolis Quarterly Review, 28 (1), pp 2 –15.

    • Search Google Scholar
    • Export Citation

  • Quah D. (1997) “Empirics for Growth and Distribution: stratification, polarisation, and convergence clubs” Journal of Economic Growth, 2.

    • Search Google Scholar
    • Export Citation

  • Rajan R. and L. Zingales (1998) “Financial Dependence and Economic Growth,” American Economic Review, Vol. 88, No 3, pp 559 –586, (June).

    • Search Google Scholar
    • Export Citation

  • Romer P. (1990) “Endogenous Technological Change,” Journal of Political Economy, 98, October, part II, S71 –S102.

  • Sala-i-Martin X., Doppelhoffer G. and R. Miller (2003) “Determinants of Long-term Growth: A Bayesian Averaging of Classical Estimates (BACE) Approach,” Columbia University, mimeo.

    • Search Google Scholar
    • Export Citation

  • Sapir A. (2005) “Globalisation and the Reform of the European Social Models,” September, available online at http://www.bruegel.org/Repositories/Documents/publications/working_papers /EN_SapirPaper080905.pdf

    • Search Google Scholar
    • Export Citation

  • SVIMEZ (2005) Rapporto 2005 sull’economia del mezzogiorno, Il Mulino, Bologna,

  • Vamvakidis A. (2002) “Regional Wage Differentiation and Wage Bargaining Systems: The Case of Italy” pp 28 –51.

  • Vamvakidis A. (2003) “Regional Convergence in Italy: 1960–2002,” Italy Selected Issues, pp. 31 –54

  • Young A., Higgins M. and D. Levy (2003) “Sigma Convergence versus Beta Convergence: Evidence from Country-level Data,” Department of Economics, Emory University, Emory Economics 0316, September.

    • Search Google Scholar
    • Export Citation
1

Prepared by Bogdan Lissovolik (EUR)

2

See the discussion of and references to “Europessimism” in Blanchard (2004), as well as his relatively sanguine interpretation of Europe’s challenges.

3

For example, earlier studies (i.e., Decressin (2000)) emphasized labor market rigidities, but ongoing progress on this front has so far done little to improve Italy’s overall growth record.

4

The literature on general determinants of economic growth has recently turned to these short-to-medium-term timeframes (see Cerra and Chaman Saxena (2005) and Jones and Olken (2005)).

5

P-convergence essentially refers to the negative relationship of per capita real economic growth with the lagged initial level of real per capita income for a cross-section of countries or regions. σ-convergence denotes a decline in the cross-sectional standard deviation for the log of per capita income. See Barro and Sala-i-Martin (1995) for details.

6

They did find, however, convergence with respect to labor productivity.

7

The PPP-adjusted per capita income has been used instead of the more common real per capita income, given that the updated NUTS-2 level data are available for a longer period of time. Both measures have been used in practice (Vamvakidis (2003)). In any case, results with these two measures of convergence are similar for those periods for which the available data overlap.

8

The results would be, however, somewhat different for σ- convergence, which holds for EU-25, but there is no convergence or divergence for EU-14. However, this is consistent with the result that β-convergence is necessary but not sufficient for σ- convergence (see Young and others (2003)).

9

This may not be surprising given that much of the recent drive for European integration has occurred at the EU-wide level.

10

The usefulness of available region-specific data at the NUTS-2 level of disaggregation (beyond the concept of convergence) is so far limited, partly because of the lack of series for some variables and partly due to the absence of relevant deflators for others (see Appendix).

11

The contemporaneous change in the tax burden has not been included because of the clear potential for reverse causality, as slow growth may cause an increase in the tax ratio (see Appendix for more discussion of reverse causality).

12

Indeed, replacing the labor tax rate with a more general tax/GDP ratio eliminates the significance of the level of the EPL index, while the contemporaneous change remains significant.

13

Most likely, this reflects the fact that the rate of recovery - as a broad measure of institutional quality - somewhat overlaps with the PMR. Our baseline specification is however based on the PMR, given that the data on the former measure is not available prior to 2004, raising concerns about reverse causality.

14

However, De Nardis and Trau (1999), on the basis of more disaggregated analysis, find that, accounting for quality differentials, competitive pressures on Italy would be much lower than implied by simple comparisons of sectoral structures.

15

Including available region-specific data available for the NUTS-2 disaggregation (employment, unemployment, population size, human capital measures (education, etc.)) was likewise not significant.

16

At the same time, however, a single-country focus also involves a loss of some data, since many variables available for cross-country analysis (i.e., structural indices) are not compiled on a within-country basis.

17

This estimator uses first differences to eliminate region-specific effects and employs lagged values of explanatory variables as instruments to deal with endogeneity problems. The underlying assumptions are that the error term is not serially correlated and the explanatory variables are weakly exogenous.

18

However, this argument to some extent contradicts the observed Italy-wide shortfall in growth in sectors such as textiles and leather.

Same Series

Other IMF Content

Other Publishers

Cover IMF Staff Country Reports
Italy: Selected Issues
Author: International Monetary Fund
Volume/Issue:
Volume 2006: Issue 059
Publisher:
International Monetary Fund
ISBN:
9781451819885
ISSN:
1934-7685
Pages:
142
DOI:
https://doi.org/10.5089/9781451819885.002
  • View in gallery

    EU-25, real PPP GDP per capita convergence, 1995–2002

  • View in gallery

    EU-14: Product and Labor Market Indicators, 1998-2003.

  • View in gallery

    EU-5: Pairwise Products of Sectoral Symmetric Balassa Indices of EU-5 countries with China, 1997

  • View in gallery

    Italy, Asia-prone Production in Regional Manufacturing and Exports, 2001–2004.

    (in percent)

  • View in gallery

    Italy: Real GDP per Capita Convergence.

  • View in gallery

    Italy: Regional Convergence and Asia-prone Sectoral Shares, 2001–04.

    (bubbles denote Balassa sectoral specialization indices)

  • View in gallery

    Comparison of coverage of structural reforms with the IMF (2004) study.

EU-25, real PPP GDP per capita convergence, 1995–2002

EU-14: Product and Labor Market Indicators, 1998-2003.

EU-5: Pairwise Products of Sectoral Symmetric Balassa Indices of EU-5 countries with China, 1997

Italy, Asia-prone Production in Regional Manufacturing and Exports, 2001–2004.

(in percent)

Italy: Real GDP per Capita Convergence.

Italy: Regional Convergence and Asia-prone Sectoral Shares, 2001–04.

(bubbles denote Balassa sectoral specialization indices)

Comparison of coverage of structural reforms with the IMF (2004) study.