Italy: Selected Issues
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar Close
Publication Date:
13 Nov 2003
eISBN:
9781451819816
Language:
English
Keywords:
ISCR; CR; inflation differential; pension; real GDP; price level convergence; inflation expectation; euro area; price-level differential; inflation term; Output gap; Productivity; Consumer price indexes
Download PDF (1 MB)

The Selected Issues paper is focused on policies to secure strong growth and safeguard fiscal sustainability. The paper analyzes the reasons behind Italy's persistent inflation differential vis-a-vis the euro area. It reviews Italy's large regional imbalances through a catch-up in income levels and estimates a growth model using panel data for Italian regions to determine the impact of a number of factors in addition to convergence forces. It also focuses on fiscal sustainability and reviews the case for additional pension reform steps in Italy.

Abstract

The Selected Issues paper is focused on policies to secure strong growth and safeguard fiscal sustainability. The paper analyzes the reasons behind Italy's persistent inflation differential vis-a-vis the euro area. It reviews Italy's large regional imbalances through a catch-up in income levels and estimates a growth model using panel data for Italian regions to determine the impact of a number of factors in addition to convergence forces. It also focuses on fiscal sustainability and reviews the case for additional pension reform steps in Italy.

Inflation and Competitiveness1

A. Introduction

1. The steady rise in the Italian inflation differential versus the euro area over the past two years has raised concerns regarding the persistence of inflation in Italy, and its possible implications for competitiveness. After falling to zero in 1997, the (headline) inflation differential has since averaged about ½ percentage point (and was as close to 1 percentage point in the summer of 2003, also in underlying terms; Figure 1). Over the same period, Italy’s competitiveness—as measured, for example, by export market share—has been in decline.

Figure 1.
Figure 1.

Italy: Inflation Differential Versus the Euro Area, 1991-2003

(In percent)

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

2. With various data sources suggesting that Italy has a price level somewhat below that of the euro-area average, gradual convergence toward the area average could explain a persistent inflation differential. It might also be explained by only gradual adjustment from the relatively high inflation prior to monetary union. This chapter uses panel data for the euro area from monetary union in 1999 to 2002, to investigate these and other factors driving the inflation differential, examine its likely persistence, and consider the implications for Italian competitiveness.

3. In the context of monetary union, this inflation differential could present a loss of competitiveness vis-à-vis Italy’s euro-area partners—at least to the extent that the inflation differential is not driven by relative gains in productivity in the traded sector. Indeed, Italy’s real exchange rate (vis-à-vis the euro area) has appreciated and its export market shares relative to other euro-area countries have declined in recent years (see Figures 2 and 3 below).2 These losses in competitiveness may have contributed to a slower pace of economic activity.

Figure 2.
Figure 2.

Italy: Export Market Share, 1989-2003

(1989=100)

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

Figure 3.
Figure 3.

Italy: Export Market Share Relative to Germany and France, 1989-2003

(1989=100)

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

4. The rest of the chapter is structured as follows. Section B reviews possible explanations for the inflation differential—highlighting a number of both persistent and temporary factors—and summarizes some of the existing empirical evidence for Italy and the euro area more generally. Two hypotheses are introduced, first that Italy’s high inflation represents the effects of convergence starting from an initially low level of prices; and, second, that it reflects the slow adjustment of expectations to the high inflation of the past. Section C analyses these hypotheses empirically using a panel data model for the euro area. Section D considers the explanatory role played by measures of productivity or income, and considers a range of alternative estimates of the output gap as explanators of the inflation differential. Conclusions are drawn in Section E.

B. Determinants of Inflation Differentials: Persistent and Temporary Factors

5. This section provides an overview of the range of factors that could account for the inflation differential, with a brief discussion of the existing evidence.

6. To start with, Italy has recorded relatively higher inflation than the euro-area average, not just in the aggregate, but also across most broad categories of goods and services. The table below shows the Italian inflation rates and the differential versus the euro area for the 12 broad categories of the harmonized indexed consumer prices (HICP). Italy experienced higher inflation in 10 of these categories, with only alcoholic beverages and education recording lower inflation. The inflation differential was especially high for the communications category. However, even if the inflation differential in this category had been zero (from 1998 to 2002), the aggregate differential would still have been 0.34 percentage point (versus 0.4 percentage point observed for the overall HICP).

7. Italy has not suffered higher inflation than the euro area because of higher shares for those categories of the consumption basket that had higher inflation than other categories. Relative to the euro area, Italy has somewhat higher shares in the HICP basket for the categories of restaurants, hotels, clothing, and furnishings, and lower weights on housing and transport. However, taking the Italian inflation rates for the 12 broad HICP categories, but applying euro-area average weights, produces an inflation rate that is just above the actual one experienced by Italy (by 0.02 percentage point from 1998 to 2003).

8. A possible longer-run determinant of inflation differentials is price level convergence within the euro area. For traded goods and services, this would follow from arbitrage within a tightly integrated trading community, although, differences here may be quite persistent due to variation in prices of nontraded goods and services that form an important part of the final price facing consumers for traded goods and services.3 For nontraded prices, convergence would follow from gradual convergence in income and productivity over time. This may be aided by similar institutional structures, as arguably provided by the monetary union of the euro area, and the framework for trading and competition provided by the European Union.

Italy: Inflation, Inflation Differential Versus the Euro Area, and Item Weights

(1998-2002, annual average, in percent)

article image
Source: Eurostat.

9. There is evidence of price convergence within Europe, though it appears to be quite gradual. Rogers (2002) examines a set of disaggregated price data for euro-area cities from 1990 to 2001 and finds evidence of a large decline in dispersion of traded prices across Europe (to a level close to that between cities within the United States). He also finds a decline in the dispersion of nontraded prices, albeit to a lesser extent.4 Estimates from Honohan and Lane (2002) imply that if a country’s price level is 10 percent lower than the euro-area average, this would contribute between 0.3 and 0.4 percentage point to the annual inflation differential (Rogers, 2002; and Honohan and Lane, 2003). This is somewhat slower than within the United States, which has much stronger linkages across regions in several important respects, including for fiscal policy and labor mobility. Cecchetti, Mark, and Sonora (2001) estimate that across U.S. cities, the half life of convergence is about nine years—that is, a 10 percent price-level differential would contribute just over 0.5 percentage point to the annual inflation differential.

10. The various measures of the aggregate price level all suggest that Italy is currently somewhere below the European average. Data from the Economic Intelligence Unit survey of prices of comparable baskets of goods and services in different cities (used in Rogers, 2002) suggests that the price level in Italy is currently 15 percent below the European average. Aggregate Eurostat data suggest it is only 5 percent lower, while the consumption price level from the Penn World Tables (PWT) implies an intermediate figure of 10 percent.

11. One mechanism that could underlie price level convergence is the Balassa-Samuelson effect, whereby countries with lower productivity in the traded sector experience more rapid productivity growth on the path of convergence. The adjustment process leads to a higher rate of wage inflation in the economy as a whole, and hence a positive inflation differential. For Italy, however, this seems unlikely given its relatively high level of productivity.5

12. An alternative possibility is that Italy is only gradually adjusting to the large nominal depreciation of the early 1990s, which had driven the price level below the euro-area average.6 This could explain the relatively low price level, but suggests that the temporary boost to competitiveness (in spite of a possible structural trend decline in market share, discussed in the accompanying 2003 staff report on Italy) has helped to raise demand above what it otherwise would have been.

13. Convergence in income levels could also explain convergence in price levels if demand influences relative prices. For example, higher income countries might spend relatively more on nontraded or service sectors (for example, if demand is not homothetic), leading to relatively higher prices in these sectors if production here relies on a fixed factor of production. However, an examination of trends in consumption shares and inflation (relative to the euro area) across different categories of goods and services suggests that a shift in the pattern of demand has not been a factor influencing the relative prices of goods and services in a way that could affect the inflation differential.

14. A cross-country comparison of the price level in 1998 (from the PWT) and the average inflation rate from 1999 to 2002 is consistent with the price convergence effect—that is, a low price level initially is associated with a higher average inflation rate (Figures 4 and 5; and Table 1). The relationship between initial productivity levels and average inflation over this period is less clear cut—with a wide range of average inflation experienced by countries within the mid range of productivity levels in 1998 (Figures 4 and 5).

Figure 4.
Figure 4.

Italy: Initial Price Level Versus Average Inflation

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

Figure 5.
Figure 5.

Italy: Initial Productivity Versus Average Inflation

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

Table 1.

Summary Statistics for Euro-Area Countries

(In percent, unless otherwise noted; differentials are relative to the euro area)

article image

Excluding Greece.

15. Another possible explanation for relatively high inflation could be a slow convergence of expectations regarding inflation of wage and price setters within a country following the fixing of nominal exchange rates within the monetary union. Such an effect might, a priori, explain most of the inflation differentials, or it could operate in conjunction with other persistent factors. This effect would make sense for countries like Italy, Portugal, and Spain, which had relatively high average inflation for the ten years prior to meeting the convergence criteria in 1997, and have also had large (though not the highest) average inflation differentials subsequently (Figure 6). These inflation differentials could persist for some time before eventually leading to a loss in competitiveness, slowing economic activity sufficiently until people’s expectations for inflation are brought into line with the rest of the euro area.

Figure 6.
Figure 6.

Italy: Inflation of 1980s Versus Current Average Inflation

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

16. While the types of factors described above imply a gradual path of adjustment—along which inflation differentials would be expected to persist—there are a host of other factors that can contribute to inflation differentials in the short run. Countries may respond differently to common shocks. This may partly reflect variation in consumption patterns,7 or in production—for example, some countries might be more dependent on oil—or both, for example through different foreign trading partner dependencies. It may also reflect variation in the persistence of inflation due to differences in nominal rigidities in product and labor markets. The OECD (2002) summarizes this evidence, suggesting greater rigidities in Italy on both counts.8

17. Another possibility is that countries are subject to country-specific shocks. There are a number of shocks that can be incorporated into empirical analysis, including:

  • Demand and supply shocks, leading to a temporary output gap. Figure 7 shows the positive relationship between the average output gap and the average inflation rate from 1999 to 2002, with Ireland and Germany at opposite extremes;

  • Changes in indirect taxes have been large in some EU countries. Across euro-area countries, there is a small positive correlation (of 0.24) between the average change in indirect tax rates and the inflation rate over the period from 1999 to 2002. The sharp rise in indirect taxes in Italy from 1997 to 1998 (relative to the euro area) appears to have had a significant impact on inflation in 1998. Thereafter, the level of indirect taxes has remained close to that of the euro area, and is unlikely to be a factor driving the inflation differential in the future.

  • Liberalization of product markets. The OECD provides detailed comparisons regarding the extent of product market liberalization, showing Italy lagging behind many euro-area countries (OECD 2003). This study suggests that despite a low price level overall, Italy suffers from relatively high prices in key nontraded sectors including: electricity and gas—inflation has been similar, but prices remain high; postal and telecommunications—despite liberalization, prices remain high and have not declined as rapidly as elsewhere in the euro area; wholesale and retail distribution—the lowest productivity in the EU, and high markups; professional services—with above average regulation. However, the OECD indices of product market regulation are not available consistently for all these countries over time. Instead, the business regulation subindex of the Economic Freedom Index is used (see appendix for details), which is available across time and countries.

  • Three other factors were examined but were found to be insignificant in the empirical specifications examined below; and so they are not reported in detail. First, a measure of fiscal pressures—the difference of the current cyclically-adjusted fiscal surplus (in percent of potential GDP) from the average of the previous six years. This captures the possibility that government spending may fall disproportionately on nontraded goods and services, and could influence the inflation differential for these sectors, at least in the short run. Second, the nominal effective exchange rate—the lagged (log) change in the nominal effective exchange rate (the lag captures delayed pass-through). This can vary across countries because of the variation in trading partner shares, and could influence consumer prices both directly. Third, a cubic measure of the price level—to capture the possibility that convergence may be much more gradual for countries close to the average.9

Figure 7.
Figure 7.

Italy: Inflation Versus Average Output Gap

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

C. Model Estimates—Price Level Convergence Versus Inflationary Expectations

18. Having laid out some of the factors that could explain the inflation differentials within the euro area, this section quantifies their relative importance by estimating a panel regression for the inflation differential. The approach follows that of studies by the ECB (1999), Rogers (2002), and Honohan and Lane (2003) (hereafter HL).

19. The goal of this section is two-fold. First, to investigate the validity of the competing hypotheses of price level convergence versus persistence following from lagged inflationary expectations. And, second, to examine the significance of some of the other short-term variables discussed above. Price level convergence could partly capture the impact of productivity and income convergence (Balassa-Samuelson effects), and an examination of the additional explanatory power of direct measures of the level of productivity and income is left until Section D.

Panel regression model

20. The methodology follows most closely that of HL, who use multivariate panel regressions to explain inflation differentials within the euro area. Data is annual, starting from the adoption of the euro in 1999 up to 2002, and covers the original 11 members of the euro area.10 The general specification is expressed in terms of the inflation differential with respect to a reference euro-area country (see below):11

π i t − π t R = β ( Z i t − z t R ) − δ ( [ P i t − 1 − P i t − 1 * ] − [ P t − 1 R − P t − 1 R * ] ) + ϵ i t − ϵ t R ( 1 )

where: π it and πtR are the annual national and reference-country inflation rates; Pit and PtR are the national and reference-country price levels, and the same variables with stars are the national and reference-country long-run equilibrium price levels; zit and ZtR are other national and reference-country variables that influence inflation; and εit and εtR are national and reference-country shocks to inflation.

21. The key assumption is that all countries within the euro area will converge to the same price level in the long run, that is, Pit*=PtR* . This allows equation (1) to be rewritten as:

π i t − π t R = β * ( Z i t − z t R ) − δ ( P i t − 1 − P t − 1 R ) + ε i t − ε t R . ( 2 )

22. Country fixed effects are ignored because long-run price level convergence is assumed, absent persistent productivity differentials.12 For convenience, the reference-country variables (including the reference-country shock) can be grouped into a time dummy:

π i t − π t R = φ t + β * z i t − δ P i t − 1 + ε i t . ( 3 )

23. The regression analysis is conducted by excluding the reference country from the sample (replacing its influence by time dummies). The reference country could in theory be any one of the euro-area countries, but in practice it makes sense to choose a country which behaves most like the euro-area average—since countries further from the average potentially provide more valuable information about the relationship between inflation and the various explanatory variables. This happens to be Belgium, which in terms of the key variables of inflation, the price level, and the output gap, is the country closest to the respective euro-area averages. To aid interpretation of the results, contributions of the various explanatory variables to the inflation differential are expressed later in the chapter relative to euro-area averages.

24. Using this simple framework, a number of specifications are estimated, including comparing results for headline and underlying inflation (the latter excluding energy and seasonal foods); both are based on harmonized indices of consumer prices.

25. The price level variable is the consumption price level in the Penn World Tables (see the appendix for details). The equivalent Eurostat measure does not appear to be as reliable in the sense that the variation in the price level over time is very erratic, and in many cases, poorly correlated with the cumulative inflation differentials over time.

26. It is difficult to determine the significance, if any, of the effect of lagged inflationary expectations. This is because if price convergence explains the inflation differential, the price level itself will be highly correlated with the average inflation differential over recent years. To gauge the significance of the expectations effect, a ten-year moving average of inflation is used that ends six years prior to each period. In other words, for the first observation used in the regression, 1999, the inflation expectation is measured by the average of inflation from 1984 to 1993. This measure clearly identifies countries with high inflation in the 1990s prior to monetary union—namely, Italy, Portugal, and Spain (Table 1).

27. The other variables considered here are as follows:

  • GAPit is the percentage difference between actual and potential real GDP. Potential is estimated by applying the Hodrick-Prescott filter to actual output (starting with data in 1980).13 This seems more appropriate than the use of say the OECD or WEO estimates of the output gap, since these are by construction correlated with the inflation rate (through estimates of the NAIRU). However, the results are later in this chapter compared with those using other estimates of the output gap;

  • ΔTAXit is the growth rate (in percent) of indirect taxes, measured as one plus the estimated tax rate, which is approximated by indirect tax revenues as a share of private consumption;

  • The contemporaneous change in a country’s relative ranking regarding the extent of regulations affecting business sector competition—where a positive value indicates an improvement in the relative ranking. This variable is intended to capture the impact of cross-country variation in the extent of product market liberalization on the inflation differential. It is based on the raw business regulations subindex of the Economic Freedom Index (see appendix). A lack of data means that regressions containing this variable end in 2001; and

  • Lags of all these variables were also examined, but they were not significant.

The appendix contains a more detailed discussion of, and a list of sources for, the data.

Results

28. Selected results of the panel estimates are presented in Table 2, with various model specifications numbered in the columns. The major results are robust to various checks—including: the exclusion of countries one at a time from the main regressions, variation in lags for some variables (see above), extending the sample back to 1998, and truncating it at 2001. These are discussed where relevant.

Table 2.

Panel Regressions: Determinants of Euro-Area Inflation Differentials 1/

(OLS estimates, 1999-2002; 10 euro-area countries, i.e., excludes Belgium and Greece)

article image

Dependent variable: harmonized index of consumer prices (HICP). All models contain time dummies. T-statistics in brackets. Newey-West estimates of the covariance matrix are used to correct standard errors for heteroskedasticity and serial correlation.

Average inflation from t-16 to t-6.

Degrees of freedom reflect missing observations for the business regulation rank variable in 2002.

29. The findings can be summarized as follows:

  • The coefficient on indirect taxes is positive, but insignificant in model (1), and so is not included in the remaining models.14

  • The coefficient on the output gap is positive, significant and stable under most specifications. The range of estimates are tightly clustered around 0.3, implying that an output gap of 1 percent above that of the euro-area average contributes to a positive inflation differential of 0.3 percentage point. Section D will also examine the effect of using alternative measures of the output gap. Robustness tests (not shown) imply that a significant contribution to the significance of this coefficient comes from Ireland (for which the largest positive inflation differential has been associated with the largest positive output gap, but with a relative price level close to the average).

  • The coefficient on the lagged price level is negative and significant under all specifications, indicating the impact of convergence. The speed of convergence is comparable with that found by HL (other than for model (1), which includes the inflation expectations term; see below). For Italy, for example, with a price level around 10 percent below that of the euro-area average, price level convergence implies a positive inflation differential of between 0.3 and 0.4 percentage point per year (and of up to 0.6 percentage point for some models presented in Section D; see also Tables 3 and 5).

  • The role of inflation expectations (proxied by the lagged inflation term) can be seen by comparing the results of models (1), (2), and (3). In model (1) its coefficient was negative and significant, while that for the price level is larger in absolute terms than for model (2), and the fit of the regression is greater than for models (2) and (3). This negative coefficient on lagged inflation is odd and probably reflects the high correlation between the two variables (of -0.83, between 1998 and 2002). The negative coefficient implies that other things equal, higher inflation in the past implies lower inflation currently. However, excluding the price level from the regression (model 3) leads the coefficient on lagged inflation to become positive (and significant), but at the expense of a much poorer fit (than model 2)—this fit can be improved, however, only by shortening the moving average period and reducing the lag of this expectations measure (not shown). In short, the measure of lagged inflation used (as a proxy for backward-looking inflation expectations) is a significant explanator of the inflation differential by itself, but adds to the regression in a way that suggests possible over fitting of the data.

  • The coefficient on the change in the (contemporaneous) business regulation ranking of a country is positive but not significant—with a probability value, however, only just above the cut-off value for the 10 percent significant level—in model (4) (it is examined only in the context of the more parsimonious specifications due to a lack of data for this variable, and hence, a degrees of freedom problem). Using the OECD measure of the output gap in place of the HP filtered version (model 6), does lead to a positive and significant coefficient (at the 10 percent level) for the business regulation rank variable. The magnitude of the coefficient implies that an improvement in a country’s ranking by one is associated with a decline in the inflation differential of about 0.1 percentage point in that same year (a one year lag of the business regulation ranking was insignificant).

  • Coefficient estimates (on the price level and the output gap) and the fit of the model are of the same order of magnitude when the all items HICP inflation is replaced on the left-hand-side of the regression by the underlying HIPP inflation.

  • Finally, an examination of the residuals from these regressions points to the Netherlands in 2001 and 2002 as being the only major outlier, with an inflation differential well above that predicted by the models. The prediction errors for Italy are in line with those of other countries, and the inflation differential for Italy was not over or under predicted over the sample as a whole.

Table 3.

Panel Regressions: Determinants of Euro-Area HICP Inflation Differentials 1/

(OLS estimates, 1999-2002; 9 euro-area countries, i.e., excludes Belgium, Greece, and Luxembourg)

article image

Dependent variable: harmonized index of consumer prices (HICP). All models contain time dummies. T-statistics in brackets. Newey-West estimates of the covariance matrix are used to correct standard errors for heteroskedasticity and serial correlation.

Average inflation from t-16 to t-6.

Degrees of freedom reflect missing observations for some countries for the traded versus nontraded measure of productivity.

Table 4.

Italy: Estimated Contributions to the Consumer Price Inflation Differential, Euro-Area Countries, 2002

(In percentage points; differential is relative to the euro area)

article image
Table 5.

Panel Regressions: Determinants of Euro-Area HICP Inflation Differentials 1/

(OLS estimates, 1999-2002; 9 euro-area countries, i.e., excludes Belgium Greece, and Luxembourg)

article image

Dependent variable: harmonized index of consumer prices (HICP). All models contain time dummies. T-statistics in brackets. Newey-West estimates of the covariance matrix are used to correct standard errors for heteroskedasticity and serial correlation.

Includes Luxembourg.

Degrees of freedom reflect missing observations for the business regulation rank variable in 2002.

D. Productivity and the Output Gap

30. The previous section demonstrated the significance of the initial price level as an explanatory variable of the inflation differential, and suggested no significant role for a country’s earlier history of inflation. This section considers the significance of measures of productivity along side that of the price level. Measures of productivity might capture more directly the Balassa-Samuelson effect than the price level itself.15 Even if this is true, the price level may still maintain a significant role given that it should also capture the situation in Italy, which has relatively high productivity, but a low price level following the only gradual adjustment to the earlier nominal depreciation.

Productivity

31. Two measures of productivity are considered: the lagged level of labor productivity (measured by the ratio of GDP, in PPP-constant 1995 dollars, to total employment);16 and the contemporaneous change in the (log of the) ratio of traded to nontraded productivity. Two points are worth noting before discussing the results of this analysis. First, Luxembourg is an extreme outlier with respect to output per worker (at almost double the euro-area average level), and so it is important to ensure that results are robust to the exclusion of Luxembourg. Second, the coefficient on productivity is dependent on the presence of the price level in the regression, so results are compared with each measure included separately, with those where both are included. Results are shown in Table 3 for the sample excluding Luxembourg (including results from the parsimonious model (2a) from Section C for comparison).17

32. The coefficient on the change in the ratio of traded to nontraded productivity is insignificant (model 8). The coefficient on the lagged productivity level is positive but insignificant in the presence of the price level term (models 8 and 9)—and leads to a slight improvement in fit (model 9 versus model 2a), The improved fit from including the productivity level comes mostly from Ireland—which has had much higher inflation than predicted by the models, and relatively high productivity—while the fit for Italy worsens. This can be seen by comparing the predicted and actual inflation differentials for models 2a and 9 shown in Table 4. Without the price level (model 10), the coefficient on the productivity level is negative (as implied by Balassa-Samuelson), but insignificant.

33. In short, the lagged level of productivity per worker does not appear to be a robust explanator of the inflation differential. Its coefficient is positive and significant when the price level term is also included—implying that countries with higher productivity have relatively higher inflation (other things equal). However, it is negative and insignificant without the price level included.

The role of the output gap

34. The fact that Italy has a relatively low price level appears at odds with the fact that it has also suffered from declining export market shares and market shares below its historical average (versus trading partners overall, though not to the same extent for euro-area trading partners). These observations can be reconciled, however, by assuming that structural shocks have reduced Italy’s export capacity, so that in spite of poor performance of late, Italy is still above its long-run equilibrium market share. (Some of these issues are discussed in more detail in the staff report for the 2003 Article IV consultation.) The key issue here is that along the path of adjustment, with the current export market share above the long-run equilibrium, there should be excess demand pressures (relative to the euro area) working to keep inflation above the euro-area average. In other words, we should expect the output gap (that is, actual output less potential output) in Italy to be higher than that of the euro area during this transition, other things equal.

35. The HP filtered estimate of the Italian output gap is, however, below that of the euro area in 1999 and 2000, and only marginally above it in 2001 and 2002 (Figure 8), implying a negative contribution initially and then only a negligible positive contribution thereafter to the inflation differential (Table 6, models 2a and 9). The OECD estimates imply a large negative contribution of the output gap to the inflation differential for Italy from 1999 to 2002 (Table 6, models 6, 2b, and 9b). These models provide a better fit of the data, but it is important to note that this is in part by construction, since the OECD gap estimates are based on estimates of the NAIRU in the first place. Models using WEO output gap estimates (2c and 9c)—also based on NAIRU estimates—provide a still better fit, with somewhat higher coefficients on the output gap (though not statistically different from models using the HP filter for the output gap). The WEO output gap estimates still imply a negative contribution to the inflation differential for 1999, 2000, and 2002.

Figure 8.
Figure 8.

Italy: Output Gap, 1997-2002

(Differential versus euro area; in percent of potential GDP)

Citation: IMF Staff Country Reports 2003, 352; 10.5089/9781451819816.002.A002

Table 6.

Italy: Estimated Contributions to the Consumer Price Inflation Differential, 2002

(In percentage points; differential is relative to the euro area)

article image

Contributions are for 2001, since data on the rank of business regulation is not available for 2002.

36. It is possible, however, that Italy’s large regional differences have important implications for aggregate supply constraints that are not well captured in the above estimates (which are based on nationwide aggregates). In particular, tight employment conditions in the North of Italy could, combined with price setting power of the North (atleast in the labor market), have contributed to higher inflation than one would have expected otherwise for the country as a whole. The wage-setting power of the northern regions is confirmed by Brunello, Lupi, and Ordine (2000), who show that wage setting in Italy depends only on the rate of unemployment in the North and Center of the country.18 Hence, it makes sense to re-estimate the output gap so as to take this into account. A simple way to do this is to use the HP filter on GDP of the North (and Center) to re-estimate the output gap for Italy. The results (not shown) are, however, very close to that for Italy overall and make little difference to the model estimates and contributions to the inflation differential.

37. An alternative is to re-estimate the output gap—following the WEO methodology (for Italy, see IMF Country Report 02/232, Chapter II)—with a new NAIRU estimated by assuming that wages depend only on unemployment in the North (defined below as all regions other than the South). That is, the NAIRUN for the North is defined as the level of unemployment above (below) which inflation is falling (rising):

D 2 log ⁡ W = − α ( U N − N A I R U N ) , ( 4 )

where: W is the nominal wage level for Italy as a whole, UN is the actual unemployment level in the North of Italy, and D is the first difference operator. Since unemployment in the South is assumed to have no influence on wages, the NAIRU for Italy overall is simply the sum of NAIRUN and the unemployment level in the South:

N A I R U = N A I R U N + U S , ( 5 )

38. The procedure used to estimate NAIRUN follows that used in IMF Country Report 02/232, Chapter II. The resulting NAIRU estimate for the country as a whole is used to re- estimate potential output for Italy, which is displayed in Figure 8 as WEO, Italy North. In contrast to the other potential output series, it leads to an output gap above the euro-area average from 1999 to 2002—by an average of 0.2 percentage point. The regression results (models 2d and 9d, Table 5) are similar to those based on the WEO output gap data. And while the contribution of the output gap to the Italian inflation differential in 2002 is similar to that of the HP filtered series (Table 6), the average contribution from the output gap (WEO, Italy North) from 1999 to 2002 was almost 0.1 percentage point.

E. Conclusions

39. Italy has recorded higher inflation than the euro area for much of the period since the beginning of monetary union in 1999. This has been accompanied by losses in competitiveness and export market shares during this period, and by generally weak output growth. To gain some insights into possible linkages behind these developments, this chapter focused on the driving forces behind the persistent inflation differential.

40. Panel regression analysis for the euro-area countries suggests that price level convergence is an important determinant of inflation differentials. And while the historical difference in inflation has some explanatory power by itself, this is much lower than it is for the price level convergence effect; moreover, the former adds at best little to the fit of a regression that already includes the price level. With Italy’s price level currently estimated around 10 percent below that of the euro-area average, the range of estimates presented in this chapter imply that the inflation differential for Italy is likely to persist at between 0.4 and 0.6 percentage point per year due to price convergence effects.

41. The empirical evidence presented in this chapter provides no firm evidence for a Balassa-Samuelson effect in Italy, consistent with the fact that its productivity level is around the euro-area average. The panel regression results showed that the level of productivity added little to a model already incorporating the price level.

42. The output gap was found to be a significant explanatory variable for the inflation differential. For Italy, the standard estimates of the output gap suggest a small negative contribution to the inflation differential on average from 1999 to 2002—which is puzzling given that the relatively low price level in Italy would suggest excess demand in Italy relative to the euro area during the path to convergence. A revised NAIRU estimate, taking into account the apparent labor-market segmentation between the North and South of Italy, seems to go some way to addressing this apparent puzzle: the revised estimates imply that Italy had probably a positive output gap on average from 1999 to 2002. Based on the panel regression results, this would have contributed around 0.1 percentage point per year to the inflation differential.

43. The results of this chapter suggest possible further real appreciation in coming years due to continued price level convergence—and this raises the ante to adopt policies to avoid potentially adverse repercussions for exports and growth. This includes addressing remaining structural weaknesses in some sectors where, indeed, price levels remain relatively high (for example in energy). The results in this chapter provide some evidence that reducing regulatory restrictions and strengthening competition could also have beneficial effects in terms of strengthening price competitiveness.

44. Finally, the results in the chapter suggest that an output gap estimate that accounts for the tight employment conditions in the North of Italy, combined with price setting power of the North (at least in labor markets), is more consistent with the relatively higher inflation in Italy than are alternative estimates of the output gap. As discussed in the staff report, increased regional wage differentiation—taking into account the higher unemployment (and lower cost of living) in the South—could thus contribute importantly to strengthening competitiveness of the South, and of Italy more generally.

APPENDIX: Data

Data is annual, covering the euro area, excluding Greece, from 1997 to 2002.

Inflation: Harmonized CPI from Eurostat (calculated from the annual index). Underlying inflation is based on the all items index excluding energy and seasonal food.

Price level: the price level of consumption from Alan Heston, Robert Summers and Bettina Aten, Penn World Table Version 6.1, Center for International Comparisons at the University of Pennsylvania (CICUP), October 2002. This is the PPP of private consumption divided by the exchange rate (national currency units per U.S. dollar); the PPP of private consumption is the ratio of the national currency value to the real value in PPP dollars. (An apparent typographical error for Spain in 2000 was corrected, (converted raw data from 63.20 to 73.20, bringing the relative price level in line with its historical past and making the change in the price level (relative to the United States) similar to that of other euro-area countries). The data have been transformed initially, so as the euro area is equal to 100 in each year. The sample is extended one year (to 2001), thereby adding 11 degrees of freedom, by assuming that the relative price level adjusts according to the inflation differential.19 This seems reasonable since innovations in this price level series are highly correlated with the inflation differential—which is not so for the alternative relative price series provided by Eurostat. As a check I also run regressions from 1999 to 2001 only (with no significant differences for the main results).

Real GDP: WEO data base. Output gap: HP filter applied to real GDP, as well as the OECD and WEO data bases.

Business regulation ranking: From the Heritage Foundation/Wall Street Journal 2003 Index of Economic Freedom (http://wwYy.heritage.org/research/features/index/2003/index.html). This assigns countries an index value ranging from 0 (the least conducive to competition) to 10 (the most conducive to competition), according to a number of objective criteria (that is, it is not survey based). The business regulations index is itself based on five component indices for: price controls; administrative conditions; time spent with government bureaucracy; the ease of starting a new business; and the extent of irregular government payments. This index requires some transformations before including it in the regression analysis. Data for the business regulations index is available only for 1995, 1999, 2000, and 2001. Linear interpolations of the indices were used to construct data for the years between 1995 and 1999. Also, the level of the index declines for all countries in the sample in 2001, yet it seems unlikely that there was an absolute decline in competitive conditions in the euro area at this time. To deal with this, I replace the level of the index, with the rank for each country (within the euro area) implied by the index.

Indirect tax rates: estimated as the ratio of indirect taxes and private final consumption expenditure, both from the OECD.

Productivity: labor productivity is the ratio of GDP in PPP constant 1995 dollars to total employment, from OECD data. Traded and nontraded labor productivity are based on the ratio of real value added to total employment in the respective sectors, from the OECD National Accounts database. The traded sector includes: agriculture; hunting; forestry; fishing; and industry (including energy, but excluding construction). The nontraded sector includes: wholesale and retail trade; repairs; hotels and restaurants; transport; financial intermediation; real estate; renting; and business activities.

Nominal effective exchange rates: WEO database.

Fiscal impulse: based on the cyclically-adjusted government primary balance, as a percent of potential GDP, from the OECD,

Real GDP and unemployment by region in Italy: SVIMEZ. Wages in Italy are based on the compensation rate for the business sector from the OECD.

Where necessary, aggregations for the euro area were done in one of three ways depending on the series. First, for indirect tax rates and labor productivity, the euro area aggregate is was the sum of the numerators divided by the sum of the denominators across. Second, for the ratio of traded to nontraded productivity, and the nominal effective exchange rate, the aggregate was the sum of percentage changes of the given series for each country, weighted by the euro value of nominal GDP in 2001. Third, for the output gap measures and the fiscal impulse, the aggregate was the sum of each series across countries, weighted by the euro value of nominal GDP in 2001.

References

  • Brunello, G., C. Lupi, and P. Ordine, 2000, “Regional Disparities and the Italian NAIRU,” Oxford Economic Papers, Vol. 52, pp, 146–177.

    • Search Google Scholar
    • Export Citation

  • Canzoneri, M., R. Cumby, B. Diba, and G. Eudey, 2001, “Productivity Trends in Europe: Implications for Real Exchange Rates, Real Interest Rates and Inflation,” Georgetown University, mimeo.

    • Search Google Scholar
    • Export Citation

  • Cecchetti, S., N. Mark, and R. Sonora, 2002, “Price Level Convergence Among United States Cities: Lessons for the European Central Bank”, International Economic Review, Vol. 43, No. 4, pp. 1081–99.

    • Search Google Scholar
    • Export Citation

  • European Central Bank, 1999, “Inflation Differentials in a Monetary Union,” in ECB Monthly Bulletin (October), pp. 35–44.

  • Engel, C., 1993,“Real Exchange Rates and Relative Prices—An Empirical Investigation” Journal of Monetary Economics, 32, pp. 507–538.

    • Search Google Scholar
    • Export Citation

  • Engel, C., and J.H. Rogers, 1996, “How Wide is the Border?,” American Economic Review, 86, pp. 1112—25.

  • Honohan, P., and P.R. Lane, 2003, Divergent Inflation Rates in EMU, Trinity Economic Papers, 2003/4, The University of Dublin, Trinity College.

    • Search Google Scholar
    • Export Citation

  • Knetter, M., 1997, “International Comparisons of Price-to-Market Behavior,” American Economic Review, 83, pp. 473–86.

  • Kravis, I.B., and R.E. Lipsey, 1978, “Price Behavior in the Light of Balance of Payments Theories,” Journal of International Economics, 8, pp. 193–246.

    • Search Google Scholar
    • Export Citation

  • Lapham, B.J., 1995, “A Dynamic General Equilibrium Analysis of Deviations from the Laws of One Price,” Journal of Economic Dynamics and Control, 19, pp. 1355–89.

    • Search Google Scholar
    • Export Citation

  • Lutz, M., 2003, “Price Convergence Under EMU? First Estimates”, University of St. Gallen, Department of Economics, Working Paper No. 2003-08.

    • Search Google Scholar
    • Export Citation

  • Micossi, S., and G.M. Milesi-Ferretti, “Real Exchange Rates and the Prices of Nontradable Goods”, in Inflation and Wage Behaviour in Europe, P. de Grauwe, S. Micossi, and G. Tullio, Clarendon Press, Oxford, pp. 209–230.

    • Search Google Scholar
    • Export Citation

  • Organization for Economic Cooperation and Development, 2002, “Inflation Persistence in the Euro Area,” in OECD Economic Outlook, Vol. 2002/2, No. 72 (December), pp. 163–71.

    • Search Google Scholar
    • Export Citation

  • Organization for Economic Cooperation and Development, 2003, 2003 Annual Review—Italy,” Economic and Development Review Committee of the Economics Department.

    • Search Google Scholar
    • Export Citation

  • Rogers, J.H., 2002, “Monetary Union, Price Level Convergence, and Inflation: How Close is Europe to the United States?,” Board of Governors of the Federal Reserve System, International Finance Discussion Papers, No. 740.

    • Search Google Scholar
    • Export Citation
1

Prepared by Christopher Kent.

2

As measured by real growth of exports of goods and nonfactor services less growth of import demand in partner countries; Fund staff estimates for 2003.

3

See for example, Engel (1993), Lapham (1995), Engel and Rogers (1996), Knetter (1997), and the seminal work in this area of Lipsey and Kravis (1978).

4

Looking at post EMU period, and a number of different macro and micro price level measures, Lutz (2003) finds no evidence of a decline in price dispersion, though this period may be too short if convergence is as gradual as implied by other studies over longer periods.

5

What matters for the Balassa-Samuelson effect is the productivity growth in the traded sector (relative to the nontraded sector) compared with other countries. Canzoneri and others (2001) suggest that this is not small in Italy. However, staff estimates from OECD national accounts data (see the appendix for a description of the data), show that relative productivity in the traded and nontraded sectors rose by slightly less than in the euro area overall, which implies a small negative contribution to Italy’s inflation differential.

6

In 1991, Italy’s price level was only 1 percent below the euro-area average according to the measure from the Penn World Tables.

7

Though this is not the case for Italy, as explained above.

8

With only a few years in the sample (see below), it is difficult to test for differences in the dynamic response of inflation to common shocks.

9

The result on fiscal pressure is consistent with that of Honohan and Lane (2003) and Rogers (2002), while that for the nominal effective exchange rate was in contrast to the finding of statistical significance by Honohan and Lane.

10

There are at least two reasons to restrict the period of investigation to post EMU. The first is to avoid the earlier period of flexible exchange rates. The second is to avoid 1997 since this was the year in which the social partners were actively working to meet the Maastricht criteria for convergence of inflation rates-in spite of forces (such as price level convergence) that may have been working to maintain considerable inflation differentials. When the regressions are extended back to 1998, it did not alter the main findings of this chapter.

11

Even if the inflation differential was calculated relative to the euro area, one country would need to be droped from the panel regression, since the euro area itself is constructed as the (weighted) average of each of its member countries.

12

This representation is like an error correction model with the lagged price level terms acting as the long-run cointegrating relationship. But given the short time period for estimation, this cannot be tested. The significance of the coefficient on the price level differential—which measures the speed of convergence—would indicate the existence of a cointegrating relationship between the (nonstationary) price level variables. Estimates below show that the coefficient on this is negative and significant. It is insignificant, however, when the model is estimated instead with fixed effects (results not reported). Given the very gradual rate of convergence and the short sample period, this is perhaps not surprising, and implies that most of the variation in the inflation differentia] appears to come from cross- sectional variation in the price level, rather than variation in the price level over time.

13

Extending the sample period to 2008 using WED projections did not alter the results.

14

Its value increases (to 0.006; and is significant at the 10 percent level) if Luxembourg is excluded from the regression (results not shown), but it does not improve the fit of the regression, and is not robust to the inclusion or exclusion of other explanatory variables. Luxembourg experienced a large rise in indirect tax rates in 1999 and 2000 without an especially high inflation differential in those years.

15

Also, the price level used may be an imprecise measure of the true price level. The true price level may be better captured by a measure of the level of real productivity per worker (working also as a proxy for real income, and, therefore, capturing possible demand and supply-side effects).

16

Honohan and Lane (2003) justify including a measure of the level and change in aggregate productivity since these should affect the long-run equilibrium price level. However, this makes little sense since they assume convergence of prices in the long run to justify the estimation methodology in the first place.

17

Estimates including Luxembourg show that the coefficient on the change in the ratio of traded to nontraded productivity is insignificant, as is the coefficient on the lagged level of productivity.

18

See also, IMF Country Report 02/232, Chapter III.

19

The model itself has an error correction form, where the long-run equilibrium is the difference between each country’s price level and that of the euro-area average (with a coefficient of one due to the assumption of price level convergence). Hence, updating the price level in this way with the inflation differential is akin to updating the long-run equilibrium.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Staff Country Reports
Italy: Selected Issues
Author:
International Monetary Fund
Volume/Issue:
Volume 2003: Issue 352
Publisher:
International Monetary Fund
ISBN:
9781451819816
ISSN:
1934-7685
Pages:
87
DOI:
https://doi.org/10.5089/9781451819816.002
  • Figure 1.
    View in gallery
    Figure 1.

    Italy: Inflation Differential Versus the Euro Area, 1991-2003

    (In percent)

  • Figure 2.
    View in gallery
    Figure 2.

    Italy: Export Market Share, 1989-2003

    (1989=100)

  • Figure 3.
    View in gallery
    Figure 3.

    Italy: Export Market Share Relative to Germany and France, 1989-2003

    (1989=100)

  • Figure 4.
    View in gallery
    Figure 4.

    Italy: Initial Price Level Versus Average Inflation

  • Figure 5.
    View in gallery
    Figure 5.

    Italy: Initial Productivity Versus Average Inflation

  • Figure 6.
    View in gallery
    Figure 6.

    Italy: Inflation of 1980s Versus Current Average Inflation

  • Figure 7.
    View in gallery
    Figure 7.

    Italy: Inflation Versus Average Output Gap

  • Figure 8.
    View in gallery
    Figure 8.

    Italy: Output Gap, 1997-2002

    (Differential versus euro area; in percent of potential GDP)

Figure 1.

Italy: Inflation Differential Versus the Euro Area, 1991-2003

(In percent)
Figure 2.

Italy: Export Market Share, 1989-2003

(1989=100)
Figure 3.

Italy: Export Market Share Relative to Germany and France, 1989-2003

(1989=100)
Figure 4.

Italy: Initial Price Level Versus Average Inflation
Figure 5.

Italy: Initial Productivity Versus Average Inflation
Figure 6.

Italy: Inflation of 1980s Versus Current Average Inflation
Figure 7.

Italy: Inflation Versus Average Output Gap
Figure 8.

Italy: Output Gap, 1997-2002

(Differential versus euro area; in percent of potential GDP)